778341 - Please create a Nightly update channel for B2G full-over-the-air updates

Status: RESOLVED FIXED

(Release Engineering :: General, defect, P1)

Description

[Lukas Blakk [:lsblakk] use ?needinfo]

The B2G update service will need to be able to ping for updates and receive updates from this channel - "nightly" same as we use for Desktop/Mobile.

Comment 1

[Nick Thomas [:nthomas] (UTC+12)]

Are we going to have Nightly/Aurora/Beta/Release builds of b2g so it's a clear fit with the existing channel names for desktop/mobile ? Presumably this is dependent on having RelEng created b2g builds (at least nightlies), and then it's a matter of making sure we set the channel in-build and publish snippets to the right place.

Comment 2

[Lukas Blakk [:lsblakk] use ?needinfo]

I believe we will have Nightly/Beta/Release builds/channels (skipping Aurora for B2G since it won't be particularly useful in that product's cycles). We're still hammering out the details of builds but should have more information within the next few weeks.

Comment 3

[Andrew Overholt [:overholt]]

John, who should own this bug?

Comment 4

[Marshall Culpepper [:marshall_law]]

Will there be separate channels for FOTA vs the standard Gecko update channels? (Are there different bugs to deal with that?)

Comment 5

[Nick Thomas [:nthomas] (UTC+12)]

The current nightly builds are being done by ateam rather than RelEng, and I saw a mention of them starting to get nightly updates going. CCing jgriffin so he's aware of this bug, and we'll need to make sure this persists as builds move onto the RelEng infra.

Comment 6

[Dietrich Ayala (:dietrich)]

John, what's the status of this bug?

Comment 7

[Robert Kaiser]

(In reply to Dietrich Ayala (:dietrich) from comment #6)
> John, what's the status of this bug?

Dietrich, as builds don't come from RelEng right now but from A-Team, we are tracking this for their builds in bug 790351 right now. I guess we'll come back to here once we have builds com from RelEng.

Comment 8

[JP Rosevear [:jpr]]

bug 790351 is marked as closed and C2 is supposed to be the milestone for switching over to releng builds.  John, will this be ready for Monday?

Comment 9

[Aki Sasaki (not active)]

This is being solved in bug 816275, for regular nightly updates.

Specifically re: FOTA, aiui, the process could be:

* the channel will remain the same
* creation of the FOTA update is manual, and IIRC jgriffin and others are going to stay in charge of this until the process is hammered out.  My memory here specifically is fuzzy, however.
* we can point our regular non-FOTA nightly update upload to a different directory
* we can move the manually created FOTA update bits into the nightly channel directory
* wait (tell everyone to update)
* point our regular non-FOTA nightly update upload to the regular directory
* anyone who missed updating will need to reflash

Comment 10

[Lukas Blakk [:lsblakk] use ?needinfo]

We had proposed at one point having a version-bumped dir structure to deal with FOTAs - that would help with keep people on a linear path without having to re-flash if people missed a FOTA

So if we took the proposal from comment 9 and adjusted it would look like:

* channel remains the same
* jgriffin/marshall_law create the manual FOTA update
* point to a new dir for the FOTA update by providing an OTA update with an update url change
* users get that that update (FOTA) which also replaces the update url with a new version-bumped dir in the channel directory

That way people who miss updates are never offered future updates - this may or may not lead to re-flashing but doesn't *require* it, technically they should be able to apply the updates between them and the latest bits.

Comment 11

[Aki Sasaki (not active)]

I think in either scenario, we're set here once bug 816275 is fixed and we decide on a process.  (Comment 10 sounds better)

Comment 12

[Aki Sasaki (not active)]

Bug 816275 is resolved fixed.
Closing per comment 9 through comment 11.

Comment 13

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

(In reply to Aki Sasaki [:aki] from comment #12)
> Bug 816275 is resolved fixed.
> Closing per comment 9 through comment 11.

Thanks Aki.